Observation API

POST /api/v1/observe/assess

Submit text for behavioral pattern assessment.

Request:

{
  "text": "The agent output to analyze...",
  "agentId": "uuid",
  "runId": "uuid"
}

Response:

{
  "analysis": [
    {
      "patternId": "BP-001",
      "patternName": "Inference Over Execution",
      "detected": true,
      "confidence": 0.7,
      "severity": "P1",
      "matches": ["my understanding", "I'll assume"],
      "detectionLayer": "regex"
    },
    {
      "patternId": "BP-002",
      "patternName": "False Blocker Reporting",
      "detected": false,
      "confidence": 0,
      "severity": "P2",
      "matches": [],
      "detectionLayer": "regex"
    }
  ],
  "summary": {
    "patternsDetected": 1,
    "patternsClean": 7,
    "highestSeverity": "P1"
  },
  "storedObservations": 1
}

All 8 patterns (BP-001 through BP-008) are evaluated. Detected patterns are stored as observations associated with the agent and run.

Detection Layers

Layer	Status	Description
regex	Active	Keyword and phrase pattern matching
semantic	Post-MVP	LLM-based behavioral meaning analysis
context-aware	Post-MVP	Cross-reference with trace data

GET /api/v1/observe/:agentId

Get behavioral observations for an agent.

Response:

{
  "observations": [
    {
      "id": "uuid",
      "patternId": "BP-001",
      "confidence": "0.7",
      "severity": "P1",
      "detectionMethod": "regex",
      "evidence": {},
      "createdAt": "2026-03-01T00:00:00Z"
    }
  ]
}